Fix for IllegalStateException when stopping re-satrting an OSGI plugin with Jersey resources

[victoralfaro-dotcms]

Problem Statement

User Story

As a software engineer, I want a fix for an issue related to OSGI plugins. When stopping or undeploying an active OSGI plugin with a Jersey resource, dotCMS stops working and requires a full restart. This occurs due to a call to RestServiceUtil.removeResource(clazz) that throws an IllegalStateException.

Acceptance Criteria

• A fix is implemented to prevent RestServiceUtil.removeResource(clazz) from causing an IllegalStateException when stopping or undeploying OSGI plugins.
• Validate that dotCMS continues to operate without requiring a restart after stopping or undeploying an OSGI plugin with a Jersey resource.
• Comprehensive unit and integration tests ensure the fix does not introduce regressions.
• Proper documentation and changelog entry are added to reflect the fix.

dotCMS Version

master

Proposed Objective

Customer Support

Proposed Priority

Priority 1 - Show Stopper

External Links

• Stacktrace: https://files.slack.com/files-pri/T028R2630-F089LTU3V2P/plugin_stacktrace.java

Assumptions & Initiation Needs

• The issue is reproducible in the latest master branch.
• The CDI integration changes are confirmed as the root cause.
• Collaboration with the QA team to validate scenarios and outcomes.

Quality Assurance Notes & Workarounds

• Ensure the issue is consistently reproducible before applying the fix.
• Use test OSGI plugins to simulate the stopping/undeploying process for verification.
• Verify the fix does not impact other CDI-related functionality in dotCMS.

Sub-Tasks & Estimates

1. Investigate the root cause of the IllegalStateException (3 hours).
2. Implement a fix to handle the lifecycle of Jersey resources during plugin stopping (6 hours).
3. Create unit tests to verify the fix (4 hours).
4. Perform integration testing with OSGI plugins (6 hours).
5. Update documentation and changelog (2 hours).

Steps to Reproduce

• Start dotCMS
• Go to plugin portlet
• Upload a plugin jar that includes a Jersey resource in its code, (e.g. user registration)
• Once started make sure it runs
• Stop it and notice how dotCMS instance becomes unstable until you need to restart it

dotCMS Version

25.01.15

Proposed Objective

Code Maintenance

Proposed Priority

Priority 2 - Important

Link to Support ticket:

https://dotcms.freshdesk.com/a/tickets/29269

External Links... Slack Conversations, Support Tickets, Figma Designs, etc.

No response

Assumptions & Initiation Needs

No response

Quality Assurance Notes & Workarounds

No response

Sub-Tasks & Estimates

No response

[spbolton]

Note I think we may have been seeing a similar error message with previous version of Jersey and although we have an exception trace it didn't impact the whole server after restart of the context, The error message seemed to occur from an internal Jersey call that forms part of the request to restart. Unfortunately, we are restricted in bumping jersey up to a higher version without upgrading to tomcat 10

We have someone else here with the same message on an reload of the Jersey context
https://stackoverflow.com/questions/70792820/servicelocatorimpl-shutdown-issue-in-jersey-2-34

I want to raise this as the issue the customer is having with the server not working after the restart may not be directly related to the exception. We probably need to find the last version removing the plugin breaks the server and we may find the error message is still there, but we can then discover the real cause of the issue unless that version does no longer throw the exception. We should also look at any other indications of a problem in the log around this message

dotcms-1      | 1. java.lang.IllegalStateException: ServiceLocatorImpl(__HK2_Generated_5,5,196525911) has been shut down

[fabrizzio-dotCMS]

What OSGI resource plugin can I use to reproduce this issue?

[github-actions]

PRs:

• fix(CDI) IllegalStateException restarting an OSGI-plugin Refs:#31185 #31259
• fix(SAML) Revert bad fix Refs: #31185 #31301
• fix(CDI) Bug in Jersey that continues to inject components using an already shutdown ServiceLocator Refs#31185 #31310

[rjvelazco]

Failed IQA

• Docker: [dotcms/dotcms:trunk_ 216ce44]

500 after re-uploading a plugin

1. Go to Plugins
2. Upload a Plugin
3. Check the plugin is working
4. Stop and Undeploy
5. Try to upload the same plugin
6. See the error

Note: It can be a plugin issue

Video

IQA-OSGI-PART-1.mov

Unable to stop/undeploy a plugin

1. Go to Plugins
2. Upload a Plugin with this name: rest-endpoint-example-0.2 (1).jar, notice that we are using white spaces and ().
3. Attempt to stop/undeploy the plugin

Video

IAQ-OSGI-PART-2.mov

[fabrizzio-dotCMS]

Looks like names with spaces are troublesome but that's another issue. Please create another card for that and let's focus on the reload stop problem we had. Thanks!

[fabrizzio-dotCMS]

Second issue will be handled here: #31348

[rjvelazco]

After a conversation with @fabrizzio-dotCMS and due to support requirements, I will create a ticket for each issue found and move this ticket to done.

cc: @nollymar @fabrizzio-dotCMS

[erickgonzalez]

Only for 24.12 LTS